1

I'm running Ubuntu 17.10 on an HP laptop. It is running normally and stable however recently I am having problems where systemd-journald starts using 100% of one CPU causing it to overheat and shut down.

I've linked the problem to a java application (Pentaho application) which when opened triggers something in systemd-journald to go crazy. I've been using this application for years without problem (and without updates) so this looks to be more of an issue with systemd. Closing the application does not stop systemd so it continues to overheat and shutdown. Another Pentaho application does not cause the same issue.

Research shows plenty of issues with high CPU from journald, but I cannot find a solution to my issue.

Is it possible to exclude an application from systemd logging? Or some other configuration or setting I could change to stop logging?

Dave A
  • 83

2 Answers2

0

If your system is shutting down from overheating you should immediately start using Intel Powerclamp. It comes when installing thermald and installing tlp is also highly recommended.

The advantage of slowing down your CPU when it is about to overheat and shutdown is it gives you a chance to analyze what is happening at that time without it shutting off completely.

The other advantage of slowing down your CPU is avoiding possible damage to a chip that costs $478 (or whatever) not including labor to replace it.

You can read more about all three here: Stop cpu from overheating

The main concern I've had about constant logging in the past is when it grows > 1 GB when normally it's about < 1 MB before it's rolled over and compressed. A disk full due to looping errors is a pain.

WinEunuuchs2Unix
  • 105,762
0

CPU burnout is very, very bad.

I would recommend downloading systems monitor. If you have synaptic installed you can do this with ease this. This would have helped to handle the high usage.

Adjusting nice value and at last resort killing offending process is better then a burnt out CPU. Of course you can handle the nice value with the renice command but the flexibility of system monitor is, IMHO, the better option

When you need to find a solution quickly, you can stop the logging service in systemctl or even disable it.

systemctl list-unit-files

will help you find the culprit and then:

systemctl stop <process>

where process is the offending name of unit you found from the first command.

NotTheDr01ds
  • 22,082
bit
  • 1